next up previous
Next: Teamwork Structure Up: Task DecompositionDynamic Role Previous: RoboCup-97

Discussion and Related Work

 

This article has presented a team member agent architecture appropriate for PTS domains. While the implementation described here is in robotic soccer, it is easy to see how the architecture would apply to other sports, such as American football. In that case, the synchronization opportunities are more frequent, but formations can change during the course of a game, roles are defined with some flexibility so that agents can adjust to opponent behaviors on the fly, and agents must communicate efficiently both between plays on a drive and during plays.

There are several other examples of non-sports-related PTS domains. Having successfully developed and deployed an autonomous spacecraft [27], NASA is now interested in multi-spacecraft missions, or constellations [30]. Since spacecraft pointing constraints limit the communication both between the spacecraft and ground control and among the spacecraft, the spacecraft must be able to act autonomously while still working towards the constellations overall goal. Using interferometry missions--in which several spacecraft coordinate parts of a powerful imaging instrument to view distant objects--as an example, the locker-room agreement could be used to define several formations to be used for viewing objects that are at various distances or in different parts of the sky. Depending on the relative locations of these objects, the various spacecraft might interchange roles as they image different objects.

Search and rescue scenarios could also be formulated as PTS domains. If several robotic agents are trying to locate victims in a remote disaster sight, they may have to act quickly and autonomously. Nonetheless, before beginning the search, they could define several formations corresponding to different geographical areas of focusing their search. Within these formations, agents would need to be assigned flexible roles given that the precise situation may not be known or may change unexpectedly. The agents might also agree, as part of their locker-room agreement to switch formations either after a certain time or as a result of some limited communication, perhaps from a unit captain.

Other PTS domains that could be applications for the team member agent architecture are hospital/factory maintenance [9] and battlefield combat [39]. Network packet routing [4] could also be formulated as a PTS domain if the network nodes are permitted to freely use network bandwidth during periods of otherwise low usage. They could then exchange policies and feedback with regards to network performance.

The remainder of this section summarizes the previous work most closely related to the teamwork structure and communication paradigm as presented in this article and gives an overview of previous research in the robotic soccer domain.





next up previous
Next: Teamwork Structure Up: Task DecompositionDynamic Role Previous: RoboCup-97



Peter Stone
Thu Dec 17 15:26:44 EST 1998